RESUMO
Tuberculosis (TB) is a public health challenge globally, and molecular testing is recommended to expedite diagnosis. Concerns that Xpert MTB/RIF assay (Xpert) may be less sensitive when testing paucibacillary samples led to the development of the Xpert MTB/RIF Ultra assay (Ultra). We evaluated the performance of Ultra against Xpert on clinical samples sent to the national reference laboratory in Singapore. In total, 149 samples collected between January 2019 and November 2020 were analysed. Mycobacterium tuberculosis complex (MTBC) was isolated from 55 cultures. Using culture as the reference standard, Ultra demonstrated higher sensitivity (96.4% vs 85.5%) and marginally lower specificity (88.3% vs 89.4%) compared to Xpert in the full cohort. When considering only paucibacillary specimens such as extrapulmonary and smear-negative samples, similar results were obtained. Reclassifying Ultra trace results (low levels of MTB are detected but no rifampicin resistant result is detected) as negative in the full cohort led to a decrease in sensitivity by 10.9% and a marginal increase in specificity by 1.1%. In instances of low bacillary load, Ultra also identified rifampicin resistance more accurately than Xpert, when corroborated against other methods such as broth microdilution, line probe assay and whole genome sequencing (WGS). One isolate tested rifampicin-resistant using Xpert and Ultra, but was phenotypically susceptible and WGS demonstrated the presence of the silent mutation Thr444Thr. Ultra is more sensitive than Xpert in the detection of MTBC and rifampicin resistance in our local setting. Nevertheless, the results of molecular testing should still be correlated with phenotypic studies.
